Compiling Machida fails #1925
Comments
|
Hi @avinash010. Sorry for have run into an issue. It appears you first ran This bit: Indicates it wasn't a build that follows the installation directions. I'm interested what steps you took so I can reproduce and figure out what the problem is. I suspect that you run things should work for you. |
|
Hi @SeanTAllen . Thanks for helping me out here I tried the steps mentioned in the https://docs.wallaroolabs.com/book/getting-started/linux-setup.html I also tried the steps suggested to rectify the problem by running |
|
hi @avinash010 could you do the |
|
hi @SeanTAllen |
|
Thanks @avinash010 |
|
@avinash010 on this end @slfritchie will be looking into this. he might have questions for you if he isn't able to reproduce. |
|
I can't replicate the problem using a Xenial 64-bit VM plus the Pony |
|
fine @SeanTAllen. @slfritchie : Thanks for the update |
|
Very odd: I can reproduce it if using an m5.large EC2 instance in Oregon using the AMI After the "Stack dump:" message appeared, the |
|
I'm sorry, but I cannot recommend using the m5.large instance type. After compiling the Pony compiler myself on an EC2 m5.large instance in US West Oregon, the compiler does not crash ... but it also has not finished after 26 minutes of allegedly running at 100% CPU. If I use an m4.large instance in US West Oregon with the same AMI, plus the same |
|
The same bad behavior appears with |
|
Thanks to research by @dipinhora ... the Meanwhile, I've started a conversation with the Pony developers for work-arounds and a fix. Current conjecture is that this is an LLVM bug that is tickled by AVX512 support by the CPUs. |
Prior to this commit, we could accidentally trigger llvm bug 30542 (https://bugs.llvm.org/show_bug.cgi?id=30542) on a cpu with the `avx512f` feature. See WallarooLabs/wally#1925 for one particular example of this occurring. This commit automagically disables `avx512f` if the LLVM in use is below version 5.0.0 to work around the issue.
Prior to this commit, we could accidentally trigger llvm bug 30542 (https://bugs.llvm.org/show_bug.cgi?id=30542) on a cpu with the `avx512f` feature. See WallarooLabs/wally#1925 for one particular example of this occurring. This commit automagically disables `avx512f` if the LLVM in use is below version 5.0.0 to work around the issue.
|
@slfritchie : Thanks for the update. Will try to set up Wallaroo in some other AWS instance type |
Prior to this commit, we could accidentally trigger llvm bug 30542 (https://bugs.llvm.org/show_bug.cgi?id=30542) on a cpu with the `avx512f` feature. See WallarooLabs/wally#1925 for one particular example of this occurring. This commit automagically disables `avx512f` if the LLVM in use is below version 5.0.0 to work around the issue.
We recommend that all users use LLVM 3.9.1. However, there is a bug that in LLVM version prior to LLVM 5 that can cause the pony compiler to crash during optimization. As seen in issue #1925, this is triggered by the usage of the axv512f optimization. This patch turns off axv512f when building Wallaroo. Closes #1925
We recommend that all users use LLVM 3.9.1. However, there is a bug that in LLVM version prior to LLVM 5 that can cause the pony compiler to crash during optimization. As seen in issue #1925, this is triggered by the usage of the axv512f optimization. This patch turns off axv512f when building Wallaroo. Closes #1925
|
I've opened a patch for this. Thanks for @dipinhora and @slfritchie for doing all the hard work. The patch will go out in our 0.3.3 release that will be happening this week (quite possibly today). |
|
@avinash010 a fix for this went out with 0.3.3 today. you should be good now. thanks you for patience. happy wallarooing! |
|
I am able to install wallaroo in my m5 instance type box. Thanks @SeanTAllen @dipinhora and @slfritchie |
|
That's awesome @avinash010. If you run have any questions, you can find us on IRC and our user mailing list. |
I am getting the below error message when i try to compile machinda. I am trying to set up Wallaroo in my Linux box. Since i was not able to reopen the previous(#1876) issue i am creating a new issue.
I am using a Linux AWS box (Instance Type: m5.large)
Distributor ID: Ubuntu
Description: Ubuntu 16.04.3 LTS
Release: 16.04
Codename: xenial
Terminal Output
The text was updated successfully, but these errors were encountered: